Scholarlink

This paper discusses the problem of predicting the occupancy of channel resources for a mobile system based on a cognitive communication system. The data used is a long-term evolution (LTE) radio environment map, a spatiotemporal grid of resources with traffic passing through the cells. The radio environment map is pre-processed and filtered for training artificial neural networks. The prediction accuracy of Kolmogorov-Arnold networks (KAN) and Long-Short-Term Memory (LSTM) is compared. A model has been developed that collects data, identifies and selects frequency zones in which data transmission occurs, trains and tests KAN and LSTM. The predictive model control algorithm is implemented in Python. The LTE simulation model and input data preparation are implemented in MatLab. The experiment results to evaluate the prediction accuracy of channel resource occupancy by KAN and LSTM are presented.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparing LSTM and KAN Models for Predicting Channel Resource Occupancy in Cognitive Radio

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • R. Bohush,
  • Y. Adamovskiy,
  • N. Naumovich

摘要

This paper discusses the problem of predicting the occupancy of channel resources for a mobile system based on a cognitive communication system. The data used is a long-term evolution (LTE) radio environment map, a spatiotemporal grid of resources with traffic passing through the cells. The radio environment map is pre-processed and filtered for training artificial neural networks. The prediction accuracy of Kolmogorov-Arnold networks (KAN) and Long-Short-Term Memory (LSTM) is compared. A model has been developed that collects data, identifies and selects frequency zones in which data transmission occurs, trains and tests KAN and LSTM. The predictive model control algorithm is implemented in Python. The LTE simulation model and input data preparation are implemented in MatLab. The experiment results to evaluate the prediction accuracy of channel resource occupancy by KAN and LSTM are presented.